LFM2-700M

MERA Created at 27.03.2026 21:56
0.243
The overall result
544
Place in the rating
Weak tasks:
603
RWSD
518
PARus
508
RCB
497
ruEthics
580
MultiQ
525
ruWorldTree
486
ruOpenBookQA
598
CheGeKa
516
ruMMLU
513
ruHateSpeech
602
ruDetox
584
ruHHH
584
ruTiE
571
USE
589
MathLogicQA
396
ruMultiAr
505
SimpleAr
451
LCS
462
BPS
529
MaMuRAMu
+16
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.078 Accuracy
RCB 0.381 / 0.361 Accuracy F1 macro
USE 0.037 Grade norm
RWSD 0.35 Accuracy
PARus 0.506 Accuracy
ruTiE 0.478 Accuracy
MultiQ 0.105 / 0.049 F1 Exact match
CheGeKa 0.004 / 0 F1 Exact match
ruModAr 0.0 Exact match
MaMuRAMu 0.367 Accuracy
ruMultiAr 0.17 Exact match
ruCodeEval 0 / 0 / 0 Pass@k
MathLogicQA 0.237 Accuracy
ruWorldTree 0.459 / 0.457 Accuracy F1 macro
ruOpenBookQA 0.508 / 0.508 Accuracy F1 macro

Evaluation on open tasks:

Go to the ratings by subcategory

The table will scroll to the left

Task name Result Metric
BPS 0.646 Accuracy
ruMMLU 0.336 Accuracy
SimpleAr 0.387 Exact match
ruHumanEval 0 / 0 / 0 Pass@k
ruHHH 0.433
ruHateSpeech 0.513
ruDetox 0.045
ruEthics
Correct God Ethical
Virtue 0.025 0.039 0.076
Law 0.051 0.067 0.065
Moral 0.051 0.062 0.068
Justice 0.021 0.028 0.083
Utilitarianism 0.023 0.028 0.088

Information about the submission

Mera version
v1.2.0
Torch Version
2.10.0
The version of the codebase
0ac3a14
CUDA version
12.8
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.57.6
The number of GPUs and their type
1 x NVIDIA A100-SXM4-80GB
Architecture
vllm

Team:

MERA

Name of the ML model:

LFM2-700M

Model size

742.5M

Model type:

Opened

SFT

Additional links:

https://docs.liquid.ai/lfm/getting-started/welcome

Architecture description:

Hybrid model with multiplicative gates and short convolutions: 10 double-gated short-range LIV convolution blocks and 6 grouped query attention (GQA) blocks.

Description of the training:

Training approach: Knowledge distillation using LFM1-7B as teacher model Very large-scale SFT on 50% downstream tasks, 50% general domains Custom DPO with length normalization and semi-online datasets Iterative model merging

Pretrain data:

Approximately 75% English, 20% multilingual, and 5% code data sourced from the web and licensed materials.

License:

lfm1.0

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=[" \n"]; \nchegeka - do_sample=false;until=[" \n"]; \nrudetox - do_sample=false;until=[" \n"]; \nrumultiar - do_sample=false;until=[" \n"]; \nuse - do_sample=false;until=[" \n","."]; \nmultiq - do_sample=false;until=[" \n"]; \nrumodar - do_sample=false;until=[" \n"]; \nruhumaneval - do_sample=true;temperature=0.6;until=[" \nclass"," \ndef"," \n#"," \nif"," \nprint"]; \nrucodeeval - do_sample=true;temperature=0.6;until=[" \nclass"," \ndef"," \n#"," \nif"," \nprint"];

The size of the context:
32768

System prompt:
Реши задачу по инструкции ниже. Не давай никаких объяснений и пояснений к своему ответу. Не пиши ничего лишнего. Пиши только то, что указано в инструкции. Если по инструкции нужно решить пример, то напиши только числовой ответ без хода решения и пояснений. Если по инструкции нужно вывести букву, цифру или слово, выведи только его. Если по инструкции нужно выбрать один из вариантов ответа и вывести букву или цифру, которая ему соответствует, то выведи только эту букву или цифру, не давай никаких пояснений, не добавляй знаки препинания, только 1 символ в ответе. Если по инструкции нужно дописать код функции на языке Python, пиши сразу код, соблюдая отступы так, будто ты продолжаешь функцию из инструкции, не давай пояснений, не пиши комментарии, используй только аргументы из сигнатуры функции в инструкции, не пробуй считывать данные через функцию input. Не извиняйся, не строй диалог. Выдавай только ответ и ничего больше.

Description of the template:
ents'] -%}{%- set ns = namespace(arguments_list=[]) -%}{%- for key, value in args.items() -%}{%- set ns.arguments_list = ns.arguments_list + [key ~ '=' ~ (value | tojson)] -%}{%- endfor -%}{%- set arguments = ns.arguments_list | join(', ') -%}{{- tool_call['function']['name'] + '(' + arguments + ')' -}}{%- if not loop.last -%}{{ ' ' }}{%- endif -%}{% else %}{{- tool_call -}}{%- endif %}{%- endfor %}{{- '</function_calls>' -}}{%- endif -%}{%- if not loop.last -%}{{- '<|im_end|>' + ' ' -}}{%- else -%}{{- eos_token -}}{%- endif -%}{%- elif message['role'] == 'environment' -%}{{- '<|im_start|>environment ' + message['content'] + '<|im_end|> ' -}}{%- elif message['role'] == 'tool' -%}{{- '<|im_start|>environment ' + message['content'] + '<|im_end|> ' -}}{%- endif -%}{%- if loop.last and add_generation_prompt -%}{{- '<|im_start|>assistant' -}}{%- endif -%}{%- endfor -%}

Ratings by subcategory

Metric: Grade Norm
Model, team 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 8_0 8_1 8_2 8_3 8_4
LFM2-700M
MERA
0 0 0.333 0.167 0 0.067 0 - 0 0 0 0 0 0 0.1 0.117 0.033 0 0 0 0 0 0 0 0 0.008 0.067 0.067 0 0.1 0.067
Model, team Honest Helpful Harmless
LFM2-700M
MERA
0.443 0.39 0.466
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Management Philosophy Prehistory Human aging Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Biology (college) Physics (college) Human Sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine learning Medical genetics Professional law PR Security studies Chemistry (школьная) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual_physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) European history Government and politics
LFM2-700M
MERA
0.311 0.392 0.428 0.474 0.366 0.488 0.476 0.395 0.37 0.323 0.237 0.278 0.31 0.38 0.322 0.399 0.38 0.313 0.233 0.328 0.269 0.339 0.25 0.301 0.286 0.4 0.307 0.361 0.42 0.27 0.46 0.603 0.356 0.485 0.37 0.35 0.25 0.319 0.298 0.325 0.359 0.243 0.476 0.207 0.361 0.301 0.373 0.152 0.241 0.341 0.29 0.418 0.318 0.353 0.39 0.436 0.358
Model, team SIM FL STA
LFM2-700M
MERA
0.478 0.361 0.403
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Managment Philosophy Pre-History Gerontology Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Bilology (college) Physics (college) Human sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine Learning Genetics Professional law PR Security Chemistry (college) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical Engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional Accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) Europe History Government and politics
LFM2-700M
MERA
0.178 0.386 0.317 0.463 0.303 0.397 0.379 0.263 0.385 0.338 0.372 0.292 0.317 0.357 0.263 0.247 0.43 0.444 0.246 0.386 0.228 0.305 0.333 0.385 0.511 0.379 0.308 0.333 0.649 0.378 0.644 0.346 0.357 0.281 0.394 0.339 0.289 0.511 0.316 0.4 0.318 0.476 0.578 0.222 0.638 0.511 0.362 0.386 0.415 0.386 0.556 0.261 0.557 0.39 0.419 0.234 0.489
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
LFM2-700M
MERA
0.025 0.051 0.051 0.021 0.023
Model, team Virtue Law Moral Justice Utilitarianism
LFM2-700M
MERA
0.039 0.067 0.062 0.028 0.028
Model, team Virtue Law Moral Justice Utilitarianism
LFM2-700M
MERA
0.076 0.065 0.068 0.083 0.088
Model, team Women Men LGBT Nationalities Migrants Other
LFM2-700M
MERA
0.491 0.514 0.529 0.432 0.429 0.607